A single molecule perspective on the functional diversity of in vitro evolved β-glucuronidase

Article abstract of DOI:10.1021/ja412379p

The mechanisms that drive the evolution of new enzyme activity have been investigated by comparing the kinetics of wild-type and in vitro evolved β-glucuronidase (GUS) at the single molecule level. Several hundred single GUS molecules were separated in large arrays of 62 500 ultrasmall reaction chambers etched into the surface of a fused silica slide to observe their individual substrate turnover rates in parallel by fluorescence microscopy. Individual GUS molecules feature long-lived but divergent activity states, and their mean activity is consistent with classic Michaelis-Menten kinetics. The large number of single molecule substrate turnover rates is representative of the activity distribution within an entire enzyme population. Partially evolved GUS displays a much broader activity distribution among individual enzyme molecules than wild-type GUS. The broader activity distribution indicates a functional division of work between individual molecules in a population of partially evolved enzymes that-as so-called generalists-are characterized by their promiscuous activity with many different substrates.

Full text of DOI:10.1021/ja412379p

Article
pubs.acs.org/JACS
A Single Molecule Perspective on the Functional Diversity of in Vitro
Evolved β‑Glucuronidase
Raphaela B. Liebherr,† Max Renner,‡ and Hans H. Gorris*,†
‡
†Institute of Analytical Chemistry, Chemo- and Biosensors and Institute of Biophysics and Physical Biochemistry, University of
Regensburg, 93040 Regensburg, Germany
S
* Supporting Information
ABSTRACT: The mechanisms that drive the evolution of new enzyme
activity have been investigated by comparing the kinetics of wild-type
and in vitro evolved β-glucuronidase (GUS) at the single molecule level.
Several hundred single GUS molecules were separated in large arrays of
62 500 ultrasmall reaction chambers etched into the surface of a fused
silica slide to observe their individual substrate turnover rates in parallel
by fluorescence microscopy. Individual GUS molecules feature long-
lived but divergent activity states, and their mean activity is consistent
with classic Michaelis−Menten kinetics. The large number of single
molecule substrate turnover rates is representative of the activity
distribution within an entire enzyme population. Partially evolved GUS
displays a much broader activity distribution among individual enzyme
molecules than wild-type GUS. The broader activity distribution
indicates a functional division of work between individual molecules in a population of partially evolved enzymes thatas
so-called generalistsare characterized by their promiscuous activity with many different substrates.
catalytic cycles can be recorded in individual chambers by
fluorescence microscopy over long periods of time. In this way,
it has been shown that individual molecules of the hydrolytic
enzyme β-galactosidase13,15,16as well as other en-
zymes12,17,18possess distinct and long-lived activity states.
The average turnover rates of individual β-galactosidase
molecules enclosed in femtoliter arrays exhibit essentially the
same dependence on the substrate concentration as the
respective bulk reactions. More importantly, these single
molecule experiments revealed a broad activity distribution
with a coefficient of variation of about 30%, which notably is
the same for all substrate concentrations investigated up to 150
μM.15
β-Galactosidase and β-glucuronidase (GUS) from E. coli
catalyze the hydrolysis of very similar glycosidic substrates.
Both enzymes are only active as homotetramers because their
active sites contain elements of two neighboring monomers as
revealed by the crystal structure (Figure 1A).19,20 The genes of
these enzymes diverged from an ancient common ancestor.21
Over the course of evolution, GUS has acquired a different
amino acid sequence and with 273 kDa22 is approximately only
half the size of β-galactosidase. Matsumura and Ellington23
showed that the native substrate specificity of wild-type GUS
can be converted to 50 million-fold higher β-galactosidase
activity by in vitro evolution within a few rounds of mutation
and screening. This study also revealed that the inversion of
INTRODUCTION
■
Analyzing the catalytic mechanisms of enzymes and their
evolution is crucial to understanding the biochemical reactions
of life. Exactly 100 years ago, the landmark work of Michaelis
and Menten provided a conceptual framework for analyzing
enzyme kinetics in bulk solution.1 The functional heterogeneity
in an enzyme population, however, cannot be resolved by
employing classic bulk reactions. Therefore, the implementa-
tion of new single molecule technologies has enabled a more
detailed description of enzyme reactions. For example, single
molecule experiments have revealed dynamic disorder in
subsequent catalytic cycles of cholesterol oxidase,2 horseradish
peroxidase,3 lipase B,4,5 β-galactosidase from Escherichia coli,6
and bovine α-trypsin.7,8
Instead of observing single substrate turnover events of only
one or a few, typically surface-immobilized enzyme molecules,
many individual enzyme molecules can be separated and
monitored free in solution in femtoliter (μm3)-sized reaction
chambers such as water-in-oil droplets,9 liposomes,10 or
femtoliter arrays etched into optical fiber bundles,11 planar
glass slides,12 or molded into PDMS.13 In particular, femtoliter
arrays etched into the surface of glass provide large numbers of
rigid and homogeneous reaction chambers for enclosing and
observing several hundred individual enzyme molecules in
parallel.14 The high degree of parallelization ensures excellent
statistics on the static heterogeneity in an enzyme population.
In addition to the single enzyme molecule, a large excess of
fluorogenic substrate is present in the femtoliter chambers.
Thus, the integrated product formation of many subsequent
Received: December 5, 2013
Published: March 31, 2014
© 2014 American Chemical Society
5949
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
Figure 1. Single molecule analysis of β-glucuronidase (GUS). (A) Crystal structure of the E. coli GUS tetramer (left) and the interface of two
monomers (center) rendered with PyMOL (PDB ID: 3k46). Four amino acid positions that are mutated in partially evolved GUS variant T509A/
D531E/S557P/N566S are highlighted in red. The right panel shows the positions of the mutated amino acids relative to the active center. (B)
Individual molecules of a conformationally heterogeneous enzyme population (E1 − En) are isolated in an array of 62 500 femtoliter chambers
etched into the surface of a fused silica slide. The array is fastened in a custom-built holder on top of an inverted fluorescence microscope and tightly
sealed by a gasket under mechanical pressure (gasket not shown). Individual GUS molecules hydrolyze the nonfluorescent substrate ReG to
fluorescent resorufin, which emits orange light. The substrate turnover of hundreds of individual GUS molecules is recorded in separate femtoliter
chambers by wide-field fluorescence microscopy. Individual trajectories of the substrate turnover are then assembled as histograms to expose the
activity distribution within the enzyme population.
GUS activity proceeds through nonspecific intermediates (or
generalists) that keep their wild-type activity but also accept a
broad range of other glycosidic substrates. After passing
through the nonspecific intermediates, a specialized mutant
GUS emerges that looses its native activity and is specific for
galactopyranoside substrates. Tawfik and colleagues24 general-
ized this observation to many other types of enzymes and
showed that it is a general principle of adaptive evolution. The
broad range of substrates catalyzed by a generalistits so-
called promiscuous activityhas been explained by a higher
diversity of conformational states.25 The high conformational
plasticity of the generalists can serve as an evolutionary starting
point for adopting new enzyme functions.
The conformational diversity of evolved enzymes has been
analyzed, e.g., by X-ray crystallography, NMR, and pre-steady-
state kinetics.26,27 Essentially all information on their
promiscuous activity has been derived from studies in bulk
solution that do not reveal how the higher conformational
diversity affects the substrate turnover of individual enzyme
molecules. Here, we have addressed this question by enclosing
wild-type β-glucuronidase and in vitro partially evolved β-
glucuronidase in large femtoliter arrays to compare their
substrate turnover rates at the single molecule level (Figure
1B).
EXPERIMENTAL SECTION
■
Protein Expression and Purification. Expression plasmids pET-
28a(+) containing sequences for N-terminally his-tagged wild-type β-
glucuronidase (GUS) or the partially evolved variant T509A/D531E/
S557P/N566S were kind gifts of Ichiro Matsumura and have been
previously described.23 The plasmids were used to transform
Escherichia coli T7 Express cells (New England Biolabs, www.neb.
com). For protein production, the transformed cells were cultivated in
1 L batches of Luria−Bertani (LB) medium at 37 °C in the presence
of kanamycin (25 μg/mL). Protein expression was induced at an
OD600 (optical density at 600 nm) of 0.4−0.6 by adding 0.5 mM of
isopropyl-β-D-thiogalactopyranosid (IPTG) to the culture medium.
The bacteria were then cultivated at 37 °C for further 16 h and
harvested by centrifugation at 4 °C for 20 min at 3000 × g. The cell
pellets were resuspended in 20 mL of nickel chelate chromatography
running buffer (50 mM sodium phosphate, pH 7.0, 300 mM NaCl, 10
mM imidazole, 5 mM β-mercaptoethanol, 1 mM EDTA) per liter
medium. Cell disruption was carried out by sonication and followed by
pelleting via centrifugation at 4 °C for 30 min at 23 000 × g. The
supernatant was loaded on a pre-equilibrated HisTrap FF crude
affinity column (GE Healthcare, www.gehealthcare.com) and eluted by
5950
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
linearly increasing the imidazole concentration of the running buffer.
The recombinant proteins were further purified by size exclusion
chromatography (SEC) using a S200 26/60 column (GE Healthcare)
pre-equilibrated in SEC running buffer (50 mM sodium phosphate,
pH 7.0, 300 mM NaCl). The purity of the resulting protein
preparations was confirmed by SDS-PAGE and protein concentrations
were determined by Bradford assay28 or absorption spectroscopy.29
Purified proteins were concentrated using Amicon Ultra 4 centrifugal
units (Millipore, www.millipore.com). Droplets of the protein solution
were snap-frozen in liquid nitrogen and stored at −80 °C. Buffer
exchanges prior to kinetic experiments were carried out using NAP-10
columns (GE Healthcare).
least 5 min on the inverted epifluorescence microscope equipped with
a precentered fiber-optical mercury illuminator (Intensilight, Nikon), a
20× objective (CFI60 Plan Apo, NA 0.75, Nikon), a filter set for
resorufin (λex = 577
10 nm, λem = 620
60 nm, Chroma
Technology, www.chroma.com) and a sensitive, high resolution (5.5
megapixels) sCMOS-camera cooled to −31 °C (Andor Technology,
www.andor.com). About 5000 femtoliter chambers were in the field of
view of the microscope. The software NIS-Elements (Nikon) was used
to control image acquisition, retrieve the fluorescence signals from
individual chambers, and process the images. The fluorescence
increase generated by individual enzyme molecules was background-
corrected by subtracting the fluorescence intensity of chambers
containing no enzyme and calibrated by determining the fluorescence
intensity of resorufin standard solutions filled into the femtoliter array
(SI Figure 2A). Enzyme kinetics was analyzed by using GraphPad
Prism 5 (www.graphpad.com).
Bulk Enzyme Experiments. The bulk activity of GUS was
determined in transparent 96-well microtiter plates (Nunc, www.
nuncbrand.com) under the same reaction conditions as in the single
molecule experiment. A bulk enzyme concentration of 36 pM equals a
single enzyme molecule in a volume of 38 fL. The substrate turnover
of various ReG concentrations was monitored on a microtiter plate
Fabrication of Femtoliter Arrays and Gasket. Femtoliter arrays
were microstructured into the surface of a fused silica wafer by
photolithography and anisotropic reactive ion etching as described in
the Supporting Information. The arrays consisted of 250 × 250 (62
500) cylindrical wells with a diameter of 4 μm and a depth of 3 μm
defining a volume of 38 μm3 (or femtoliters) as confirmed by scanning
electron microscopy (SI Figure 1). The wells were arranged in a
rectangular lattice with a pitch of 10 μm, resulting in an overall edge
length of 2.5 × 2.5 mm2. The wafer was cut into slides of 1.5 × 1.5 cm2
such that a single array was located in the middle of each slide. The
femtoliter arrays were used repeatedly and before each experiment
they were first cleaned with piranha solution (1:3 ratio of 30% H2O2
and conc. H2SO4) for 20 min, then immersed in distilled water for 10
min, sonicated for another 15 min, and finally air-dried. The femtoliter
arrays were sealed by using a polydimethylsiloxane (PDMS) gasket
(SYLGARD 184 silicone elastomer kit, Dow Corning, www.
dowcorning.com). A 0.5-mm-high layer of PDMS was cast on a
clean and smooth surface. After curing at 37 °C for 48 h, the PDMS
was cut into pieces of 5 × 5 mm2. A new piece of PDMS was used for
each experiment, cleaned with curd soap, and bidistilled water.
Buffers and Reagents for Enzyme Kinetic Experiments. Stock
solutions of the fluorogenic substrate resorufin β-D-glucuronide (5
mM) (ReG, Sigma-Aldrich, www.sigmaaldrich.com) in dimethyl
sulfoxide (DMSO) and 200 μM resorufin, sodium salt, (Invitrogen,
www.invitrogen.com) in DMSO were aliquoted and stored at −20 °C.
All dilution steps as well as the enzyme reactions were carried out at
room temperature in either GUS buffer (50 mM sodium phosphate,
pH 7.0, 5 mM β-mercaptoethanol, 1 mM EDTA) or phosphate
buffered saline (PBS: 2.7 mM KCl, 2 mM KH2PO4, 137 mM NaCl, 10
mM Na2HPO4, pH 7.4) containing 0.05 mg/mL bovine serum
albumin (BSA, Sigma-Aldrich).
Single Enzyme Molecule Experiments. For all single molecule
experiments the room temperature was maintained at 22 °C with air
conditioning. A femtoliter array was fastened in a custom-built array
holder and mounted on an inverted epi-fluorescence microscope
(Eclipse Ti-E, Nikon, www.nikoninstruments.com). The fluorogenic
substrate ReG and GUS were diluted in PBS buffer and mixed just
before an experiment was started. Varying concentrations of ReG and
approximately 1.8 pM of GUS were prepared and filled into the wells
by dispensing 5 μL of the reaction mixture on the array. The array was
covered by the PDMS gasket. A torque screwdriver was used to apply
mechanical pressure of 2.5 to 3.0 cNm on the PDMS gasket, thus
tightly sealing the array. The Poisson distribution Pμ(x) = e−μμx/x!,
where μ is the average number of enzyme molecules per chamber,
indicates the probability Pμ(x) that exactly x enzyme molecules are
enclosed in a given chamber. An enzyme concentration of 1.8 pM
enzyme in a volume of 38 fL yields a ratio of one enzyme molecule in
20 chambers (μ = 0.05). Under these conditions, most chambers
remain empty (P0.05(0) = 0.95), 5% contain only a single enzyme
molecule (P0.05(1) = 0.05), and the probability that there are more
than one single enzyme molecule in a chamber is very low (P0.05(>1) <
0.001).
reader (Fluostar Optima, bMG-Labtech, www.bmglabtech.com) (λex
=
544 nm, λem = 575 nm) and calibrated by determining the fluorescence
intensities of resorufin standard solutions (SI Figure 2B).
RESULTS AND DISCUSSION
■
Optimizing the Reaction Conditions for Detecting
Single Molecules of β-Glucuronidase (GUS). The his-
tagged GUS expressed in E. coli was highly purified by nickel
chelate chromatography and subsequent size exclusion
chromatography (SI Figure 3). Substrate saturation curves
were obtained with the chromogenic substrate para-nitrophenyl
(pNP) glucuronide because it is soluble in aqueous buffers up
to a millimolar concentration range (SI Figures 4 and 5). Wild-
type GUS has a high catalytic efficiency (kcat/KM = 8.9 × 105
s−1 M−1) as described earlier23 (SI Table 1). Additionally, we
analyzed a partially evolved GUS variant containing four amino
acid substitutions (T509A/D531E/S557P/N566S) that keeps
the original high activity for pNP glucuronide (kcat/KM = 8.3 ×
104 s−1 M−1) but also catalyzes the turnover of a broad range of
other glycosidic substrates.23
The chromogenic reaction of pNP glucuronide, however, is
not sensitive enough to report the activity of single GUS
molecules. Thus, we compared the hydrolysis of pNP
glucuronide with the respective fluorogenic substrate resorufin
β-D-glucuronide (ReG), which releases the highly fluorescent
product resorufin. The original GUS buffer system23 was
replaced by phosphate-buffered saline (PBS) containing BSA
for blocking nonspecific binding. The reaction in PBS enables a
direct comparison with the single molecule reaction of β-
galactosidase.15 The change of substrate and buffer led to a
small increase in kcat/KM (wild-type GUS: 1.6 × 106 s−1 M−1;
partially evolved GUS: 9.3 × 104 s−1 M−1). The substrate
turnover of wild-type GUS shows a typical hyperbolic
dependency in a concentration range between 12.5 and 150
μM ReG (Figure 2). Due to the limited solubility of ReG in
aqueous buffer, concentrations higher than 150 μM could not
be used. For the same reason, it was not possible to perform a
Michaelis−Menten analysis of partially evolved GUS (KM
=
Image Acquisition and Analysis. The image acquisition was
started within 2 min after mixing enzyme and substrate. The
fluorescence intensity in individual chambers of the femtoliter array
was monitored over time through the opposite face of the glass slide
by wide-field fluorescence microscopy. Images were acquired every 30
s using an exposure time of 200 ms (neutral density filter: ND 4) for at
1260 μM in the reaction with pNP glucuronide, SI Table 1).
Analyzing Wild-Type GUS at the Single Molecule
Level. The static heterogeneity of an enzyme population can
be analyzed by enclosing hundreds of single enzyme molecules
in large arrays of femtoliter-sized chambers.15 Femtoliter arrays
5951
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
chambers that contain a single GUS molecule produce a
constant amount of resorufin over time. Furthermore, the
single molecule approach in femtoliter arrays also reveals that
the substrate turnover differs among individual enzyme
molecules, which conforms to the static heterogeneity observed
earlier for β-galactosidase.15 The averaged trajectories of several
hundred individual molecules of wild-type GUS assembled
from three independent measurements are summarized in
Figure 3C. Their slope strongly depends on the substrate
concentration in the range of 12.5 to 150 μM and is consistent
with Michaelis−Menten kinetics in bulk reaction.
The turnover rates of several hundred trajectories observed
at different substrate concentrations are plotted as histograms
in Figure 4. In some histograms a very small subpopulation is
detectable, which is on average twice as active as the main
population and can be attributed to a few chambers containing
two enzyme molecules. This small subpopulation, however, can
be readily separated from the majority of single molecule
trajectories that follow a Gaussian distribution. The mean
activity calculated from single molecule trajectories is higher
than the activity of the bulk reaction because it is not possible
to completely prevent an inactive fraction of GUS (e.g.,
resulting from tetramer dissociation) in the expression and
purification steps. While a fraction of inactive enzymes would
lead to an apparent decrease in the substrate turnover rates
calculated from a bulk reaction, in femtoliter arrays only active
enzyme molecules are included in the activity calculation.14 The
substrate saturation curves of the bulk and the mean single
molecule reactions (Figure 2) yield essentially the same KM of
about 50 μM because KM is independent of the enzyme
concentration. In contrast, kcat (vmax/[E]0 = 162 s−1) calculated
from a bulk reaction is lower than kcat determined directly from
single molecule substrate turnover trajectories (283 s−1).
The activity distribution determined from the Gaussian
distribution is largely constant (CV ∼ 20%) for substrate
concentrations between 25 μM and 150 μM (Figure 4). A
slightly broader activity distribution (25%) is observed at low
substrate concentrations of 12.5 μM, which can be attributed to
Figure 2. Substrate saturation curves of wild-type GUS. Empty circles
indicate the average activity and standard deviation of three bulk
experiments (KM = 52 8 μM; kcat = 162 10 s−1), and full circles
the average activity and standard deviation of six independent single
molecule experiments (KM = 49 8 μM; kcat = 283 19 s−1).
etched into the surface of a fused silica slide were filled with 1.8
pM of GUS, which yields a single GUS molecule in every
twentieth chamber according to Poisson statistics (Figure 1).
Additionally, the chambers contained at least 12.5 μMor 3
million molecules in a volume of 38 fLof the substrate ReG.
Under these conditions, wild-type GUS hydrolyses ReG with
an average turnover rate of 50 s−1 and there is less than 1%
substrate depletion after a reaction time of 5 min. The
fluorescence intensity of the accumulating product resorufin
was recorded simultaneously by wide-field fluorescence
microscopy in more than 100 chambers that contained a single
GUS molecule. Nonspecific binding of the enzyme to the
surface of the femtoliter chambers was efficiently blocked by
adding BSA to the PBS buffer as described earlier.15
Figure 3A shows a small section of the large femtoliter array
1, 2.5, and 5 min after sealing the chambers. Four “active”
chambers that contain a single molecule of wild-type GUS and
one “inactive” chamber that contains only substrate without
enzyme are highlighted. The trajectories of these chambers are
shown in Figure 3B. The fluorescence intensity in chambers
without enzyme is constant over the whole experiment and
served for background correction (SI Table 2). By contrast,
Figure 3. Single molecule substrate turnover of wild-type GUS. (A) wild-type GUS (1.8 pM) was enclosed with 100 μM of ReG in the femtoliter
array and images were acquired every 30 s by wide-field fluorescence microscopy (SI Video). (B) Product formation in four active and one inactive
chamber (x) indicated in (A) is plotted against time. While the background fluorescence (x) is constant over time, single molecules of wild-type GUS
exhibit individually different trajectories. (C) Several hundred background-corrected trajectories of single GUS molecules are composed to ensemble
trajectories that are strongly dependent on the substrate concentration as expected from a classical Michaelis−Menten reaction.
5952
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
Figure 4. Single molecule substrate turnover distribution of wild-type GUS. Each histogram shows a representative femtoliter array experiment
recorded for 5 min at a substrate concentration of (A) 12.5 μM, (B) 25 μM, (C) 50 μM, (D) 100 μM, and (E) 150 μM. A bin time of 10 s−1 was
used for (A) and 25 s−1 for (B−E). The activity distribution in each histogram follows a Gaussian distribution. (F) Coefficient of variations (CV)
calculated from the Gaussian distributions (CV = standard deviation/mean) of six independent experiments.
Figure 5. Differences in the single molecule substrate turnover distribution between wild-type GUS and partially evolved GUS variant T509A/
D531E/S557P/N566S (100 μM substrate concentration). (A) Single molecule trajectories of partially evolved GUS. (B) The substrate turnover
histograms of partially evolved GUS (red, bin time: 10 s) and wild-type GUS (black, bin time: 25 s) both follow a Gaussian distribution. (C)
Coefficient of variations (CV) calculated from the Gaussian distributions (CV = standard deviation/mean) of six independent experiments. The
activity of partially evolved GUS (CV = 35%) is significantly more broadly distributed among individual molecules than the respective activity of
wild-type GUS (CV = 20%) (unpaired t test, p ≤ 0.0001).
low intensity trajectories and thus higher background noise. We
compared the activity distribution of GUS with β-galactosidase,
which is more than twice as active as GUS. Both enzymes
display an activity distribution that is largely independent of the
substrate concentration. Thus, the activity distribution can be
related to differences in kcat rather than KM as described
earlier.15 β-Galactosidase, however, shows a broader activity
distribution (CV = 30%) than GUS. There are several reasons
that may be implicated in the broader activity distribution of β-
galactosidase. (1) β-Galactosidase is twice as large as GUS and
thus can adopt a larger conformational space. (2) β-
Galactosidase binds two Mg2+ ions per monomer, which can
be a source of metal heterogeneity;30 GUS, however, binds no
metal ions. (3) β-Galactosidase has three catalytic functions:31
β-galactosidase not only hydrolyzes its natural substrate lactose
to galactose and glucose but to the same extent (∼50%) also
performs a transgalactosylation reaction on lactose to generate
allolactose. Allolactose binds to the lac repressor of the lacZ
gene and thus induces the lac operon and results in the
expression of β-galactosidase. As a third function, β-
galactosidase converts allolactose to the monosaccharides.
These three functions may require that β-galactosidase adopts
more conformational states. By contrast, β-glucuronides can
induce the expression of GUS directly32 and the hydrolysis of
β-glucuronides is the only known natural function of GUS.
Single Molecule Analysis of Molecular Evolution. The
in vitro evolution of enzymes proceeds via so-called generalists
that keep their original wild-type activity but also accept a broad
range of other substrates before an enzyme can acquire a high
activity and specificity for a new substrate.24 The broad range of
substrates accepted by the generalist has been explained by a
higher conformational plasticity such that one conformation
carries out the native function and an alternative conformation
is able to carry out a promiscuous function. So far, however, all
experiments showing a higher conformational plasticity of
partially evolved enzymes were performed in bulk solution
where the contribution of individual enzyme molecules remains
hidden.
5953
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
Here, we separated and analyzed several hundred individual
molecules of a partially evolved GUS variant T509A/D531E/
S557P/N566S in femtoliter chambers under the same reaction
conditions as wild-type GUS. This partially evolved GUS was
identified during the in vitro evolution of wild-type GUS toward
a higher β-galactoside activity in the second round of screening
by Matsumura et al.23 It contains four amino acid substitutions
highlighted in red in Figure 1A. Three substitutions are located
in active-site loops while D531E is located in a solvent-exposed
α-helix and was reported to have no functional effect.23 This
partially evolved GUS shows the characteristic properties of a
generalist: it is still highly active toward glucuronide substrates
but additionally hydrolyzes many other glycosidic substrates.23
Other partially evolved GUS variants isolated during the in vitro
evolution could not be investigated because their relatively low
substrate turnover rates cannot be detected in the femtoliter
array. Furthermore, it would have been interesting to analyze
the newly evolved enzyme function by using galactoside
substrates. However, even the activity of the best evolved
GUS variant (T509A/S557P/N566S/K567Q) was too low to
be detectable at the single molecule level (kcat/KM = 200 s−1
M−1 in the reaction with pNP galactoside).23
Figure 5 shows that the single molecule substrate turnover
rates of partially evolved GUS variant T509A/D531E/S557P/
N566S are long-lived, which has also been observed for
individual molecules of wild-type GUS (Figure 3). The activity
in a population of partially evolved GUS molecules, however, is
significantly more broadly distributed (CV = 35%) compared to
the wild-type activity investigated at 100 μM ReG (CV = 20%,
Figure 5B) or any other substrate concentration (Figure 4).
The broader activity distribution among individual enzyme
molecules strongly indicates that partially evolved GUS can
adopt a larger number of stable conformational states than wild-
type GUS. Frauenfelder et al.33 proposed the picture of a
rugged energy landscape for proteins, in which conformational
substates with different structural and dynamic properties are
organized in a coarse hierarchy. According to this view, there
are fast transitions between higher tiers that are separated by a
low energy barrier and slow transitions in lower tiers separated
by a high energy barrier. Even in the relatively small protein
myoglobin three distinct taxonomic substates were found.33
Protein folding is typically described as a folding funnel, where
the horizontal plane depicts the conformational freedom and
the vertical axis the free energy. A higher conformational
heterogeneity can thus be explained by a wider folding funnel
of partially evolved GUS compared to wild-type GUS. The
surface of the folding funnel of evolved GUS can keep its
rugged topography but in each local minimum more conforma-
tional substates can be accommodated. The long-lived and
broadly distributed activity states indicate a “division of work”
among the members of an enzyme population that accounts for
the functional promiscuity of the generalist. This functional
specialization of individual molecules as a motor for protein
dynamism and evolvability34,35 stands in stark contrast to the
traditional view of proteins that possess absolute functional
specificity mediated by a single fixed structure.36
catalytic functions. We suggest that the observed wider static
heterogeneity is only the “tip of the iceberg” and transitions at
lower tiers are at least equally important for the functional
heterogeneity of partially evolved GUS. Other types of single
molecule experiments, however, will be necessary to resolve
such faster transitions.
The conformational flexibility of a protein is closely related
to its thermostability because there is a trade-off between
structural stability and the flexibility required to perform a
catalytic function.37 The strong impact of a few amino acid
substitutions on the thermostability of GUS has already been
demonstrated.38,39 We analyzed changes in the secondary
structure at elevated temperatures by circular dichroism (CD,
SI Figures 6 and 7). As wild-type GUS and partially evolved
GUS differ in only four (T509A/S557P/N566S/K567Q) of
603 amino acids (Figure 1A) they adopt a very similar
secondary structure at ambient temperature. Increasing the
temperature to more than 60 °C leads to a minor change in the
secondary structure of wild-type GUS, which can be attributed
to tetramer dissociation and is accompanied by a loss of
enzymatic activity.38 The overall secondary structure, however,
remains largely unchanged up to 100 °C, thus indicating highly
thermostable monomer subunits. In contrast, partially evolved
GUS shows a typical thermal denaturation profile at temper-
atures higher than 60 °C, which is indicative of monomer
unfolding. As a lower thermostability is a sign of higher
conformational flexibility, we conclude that the more flexible
conformation of the generalist is implicated in the broader
activity distribution.
Finally, our study provides a glimpse at the evolution of
primordial enzymes. According to the “patchwork” hypothesis
first proposed by Ycas40 and Jensen,41 primordial enzymes
possessed a very broad substrate specificity that offered ancient
cells with minimal gene content a highly flexible metabolism.
Later, gene duplication provided the basis for an increasing
diversification and specialization of enzymatic activities.
Together with the recruitment of regulation mechanisms,
these steps gradually have led to the highly efficient metabolism
that we know from modern cells. We suggest that the division
of work observed among individual molecules of an enzyme
population enabled ancient cells to compensate for their low
gene content. The ability of an enzyme with a single amino acid
sequence to adopt distinct and long-lived conformational states
that catalyze various substrates allowed for an optimal use of
the limited cellular resources. Only later was the catalytic
information stored in the conformational space of individual
enzyme molecules delegated to the RNA or DNA level to avoid
disadvantageous side reactions in a more complex cellular
environment. With this transfer of information, the promiscu-
ous activity of enzymes became obsolete if not dangerous such
that mostbut not all42modern enzymes evolved to adopt a
more homogeneous activity.
CONCLUSION
■
We have demonstrated that single molecule experiments in
femtoliter arrays provide new insights not only into the catalysis
of enzymes but also how new catalytic activities evolve. β-
Glucuronidase (GUS) catalyzes a simple hydrolytic reaction
with high activity that can be readily observed at the single
molecule level and compared to the closely related enzyme β-
galactosidase. Both enzymes display long-lived but distinct
individual substrate turnover rates and their mean activities are
consistent with traditional Michaelis−Menten kinetics. Com-
It is important to note that the static heterogeneity does not
rule out a larger number of dynamic transitions at higher tiers
of the hierarchy, but these cannot be resolved with the time
resolution of our experiment. A higher conformational plasticity
could additionally lead to individual enzyme molecules that
switch between different conformational states over time
(dynamic heterogeneity)6 and thus could perform different
5954
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955

Journal of the American Chemical Society
Article
(10) Hsin, T. M.; Yeung, E. S. Angew. Chem., Int. Ed 2007, 46, 8032−
8035.
(11) Gorris, H. H.; Blicharz, T. M.; Walt, D. R. FEBS J. 2007, 274,
pared to GUS, the activity of individual β-galactosidase
molecules is more broadly distributed within the enzyme
population,15 which may be attributed to the broader range of
reactions catalyzed by β-galactosidase. A broader activity
distribution has also been observed for partially evolved GUS
variant T509A/D531E/S557P/N566S that was isolated during
the in vivo evolution of GUS into a β-galactosidase and
hydrolyzes many different substrates.23 The broad static
heterogeneity found among individual substrate turnover rates
of the generalist may be related to an ancient evolutionary
mechanism that helped primordial cells to survive in spite of a
low gene content.
5462−5470.
(12) Tan, W. H.; Yeung, E. S. Anal. Chem. 1997, 69, 4242−4248.
(13) Rondelez, Y.; Tresset, G.; Tabata, K. V.; Arata, H.; Fujita, H.;
Takeuchi, S.; Noji, H. Nat. Biotechnol. 2005, 23, 361−365.
(14) Gorris, H. H.; Walt, D. R. Angew. Chem., Int. Ed 2010, 49,
3880−3895.
(15) Rissin, D. M.; Gorris, H. H.; Walt, D. R. J. Am. Chem. Soc. 2008,
130, 5349−5353.
(16) Gorris, H. H.; Rissin, D. M.; Walt, D. R. Proc. Natl. Acad. Sci. U.
S. A. 2007, 104, 17680−17685.
(17) Gorris, H. H.; Walt, D. R. J. Am. Chem. Soc. 2009, 131, 6277−
6282.
ASSOCIATED CONTENT
* Supporting Information
■
(18) Ehrl, B. N.; Liebherr, R. B.; Gorris, H. H. Analyst 2013, 138,
S
4260−4265.
(19) Jacobson, R. H.; Zhang, X. J.; DuBose, R. F.; Matthews, B. W.
Nature 1994, 369, 761−766.
One video of single molecule substrate turnover in femtoliter
chambers, seven figures, and two tables supporting the
Experimental Section and Results and Discussion. This material
is available free of charge via the Internet at http://pubs.acs.org.
(20) Wallace, B. D.; Wang, H. W.; Lane, K. T.; Scott, J. E.; Orans, J.;
Koo, J. S.; Venkatesh, M.; Jobin, C.; Yeh, L. A.; Mani, S.; Redinbo, M.
R. Science 2010, 330, 831−835.
(21) Henrissat, B. Biochem. J. 1991, 280, 309−316.
(22) Jefferson, R. A.; Burgess, S. M.; Hirsh, D. Proc. Natl. Acad. Sci. U.
S. A. 1986, 83, 8447−8451.
AUTHOR INFORMATION
Corresponding Author
■
(23) Matsumura, I.; Ellington, A. D. J. Mol. Biol. 2001, 305, 331−339.
(24) Aharoni, A.; Gaidukov, L.; Khersonsky, O.; Gould, S. M.;
Roodveldt, C.; Tawfik, D. S. Nat. Genet. 2005, 37, 73−76.
(25) Khersonsky, O.; Tawfik, D. S. Annu. Rev. Biochem. 2010, 79,
471−505.
*Tel.: +49-941-943-4015. Fax: +49-941-943-4064. E-mail:
hans-heiner.gorris@ur.de.
Present Address
Max Renner, Division of Structural Biology, University of
Oxford, Oxford, United Kingdom.
(26) Claren, J.; Malisi, C.; Hocker, B.; Sterner, R. Proc. Natl. Acad. Sci.
U. S. A. 2009, 106, 3704−3709.
(27) Pervushin, K.; Vamvaca, K.; Vogeli, B.; Hilvert, D. Nat. Struct.
Mol. Biol. 2007, 14, 1202−1206.
Notes
The authors declare no competing financial interest.
(28) Bradford, M. M. Anal. Biochem. 1976, 72, 248−254.
(29) Pace, C. N.; Vajdos, F.; Fee, L.; Grimsley, G.; Gray, T. Protein
Sci. 1995, 4, 2411−2423.
(30) Raouda, N.; Upert, G.; Angelici, G.; Nicolet, S.; Schmidt, T.;
Schwede, T.; Creus, M. Metallomics 2014, 6, 88−95.
(31) Juers, D. H.; Matthews, B. W.; Huber, R. E. Protein Sci. 2012, 21,
1792−1807.
(32) Novel, G.; Didier-Fichet, M. L.; Stoeber, F. J. Bacteriol. 1974,
120, 89−95.
(33) Frauenfelder, H.; Sligar, S. G.; Wolynes, P. G. Science 1991, 254,
1598−1603.
(34) Tokuriki, N.; Tawfik, D. S. Science 2009, 324, 203−207.
(35) Peisajovich, S. G.; Tawfik, D. S. Nat. Methods 2007, 4, 991−994.
(36) James, L. C.; Tawfik, D. S. Trends Biochem. Sci. 2003, 28, 361−
368.
(37) Vihinen, M. Protein Eng. 1987, 1, 477−480.
(38) Flores, H.; Ellington, A. D. J. Mol. Biol. 2002, 315, 325−337.
(39) Xiong, A. S.; Peng, R. H.; Cheng, Z. M.; Li, Y.; Liu, J. G.;
Zhuang, J.; Gao, F.; Xu, F.; Qiao, Y. S.; Zhang, Z.; Chen, J. M.; Yao, Q.
H. Protein Eng. Des. Sel. 2007, 20, 319−325.
(40) Ycas, M. J. Theor. Biol. 1974, 44, 145−160.
(41) Jensen, R. A. Annu. Rev. Microbiol. 1976, 30, 409−425.
(42) O’Brien, P. J.; Herschlag, D. Chem. Biol. 1999, 6, R91−R105.
ACKNOWLEDGMENTS
■
We are indebted to Ichiro Matsumura (Emory University,
Atlanta, USA) for providing the plasmids (wild-type GUS and
variant T509A/D531E/S557P/N566S), to Prof. Reinhard
Sterner and Dr. Monika Meier (University of Regensburg,
Germany) for supporting the expression and characterization of
GUS, and to Prof. Helmut Hummel and Albert Hutterer
(Regensburg University of Applied Sciences, Germany) for
preparing microstructured fused silica slides. This work was
supported by the German Research Council (DFG grant GO
1968/3-1).
REFERENCES
■
(1) Michaelis, L.; Menten, M. L. Biochem. Z. 1913, 49, 333−369.
(2) Lu, H. P.; Xun, L.; Xie, X. S. Science 1998, 282, 1877−1882.
(3) Edman, L.; Foldes-Papp, Z.; Wennmalm, S.; Rigler, R. Chem.
̈
Phys. 1999, 247, 11−22.
(4) Flomenbom, O.; Velonia, K.; Loos, D.; Masuo, S.; Cotlet, M.;
Engelborghs, Y.; Hofkens, J.; Rowan, A. E.; Nolte, R. J.; Van der
Auweraer, M.; de Schryver, F. C.; Klafter, J. Proc. Natl. Acad. Sci. U. S.
A. 2005, 102, 2368−2372.
(5) Velonia, K.; Flomenbom, O.; Loos, D.; Masuo, S.; Cotlet, M.;
Engelborghs, Y.; Hofkens, J.; Rowan, A. E.; Klafter, J.; Nolte, R. J. M.;
de Schryver, F. C. Angew. Chem., Int. Ed. 2005, 44, 560−564.
(6) English, B. P.; Min, W.; van Oijen, A. M.; Lee, K. T.; Luo, G.;
Sun, H.; Cherayil, B. J.; Kou, S. C.; Xie, X. S. Nat. Chem. Biol. 2006, 2,
87−94.
(7) De Cremer, G.; Roeffaers, M. B.; Baruah, M.; Sliwa, M.; Sels, B.
F.; Hofkens, J.; De Vos, D. E. J. Am. Chem. Soc. 2007, 129, 15458−
15459.
(8) Terentyeva, T. G.; Engelkamp, H.; Rowan, A. E.; Komatsuzaki,
T.; Hofkens, J.; Li, C. B.; Blank, K. ACS Nano 2012, 6, 346−354.
(9) Rotman, B. Proc. Natl. Acad. Sci. U. S. A. 1961, 47, 1981−1991.
5955
dx.doi.org/10.1021/ja412379p | J. Am. Chem. Soc. 2014, 136, 5949−5955